RESUMEN

Lung cancer is the leading cause of cancer-related death in the United States. Lung adenocarcinoma (LUAD) is one of the most common subtypes of lung cancer that can be treated with resection. While resection can be curative, there is a significant risk of recurrence, which necessitates close monitoring and additional treatment planning. Traditionally, microscopic evaluation of tumor grading in resected specimens is a standard pathologic practice that informs subsequent therapy and patient management. However, this approach is labor-intensive and subject to inter-observer variability. To address the challenge of accurately predicting recurrence, we propose a deep learning-based model to predict the 5-year recurrence of LUAD in patients following surgical resection. In our model, we introduce an innovative dual-attention architecture that significantly enhances computational efficiency. Our model demonstrates excellent performance in recurrent risk stratification, achieving a hazard ratio of 2.29 (95% CI: 1.69-3.09, p < 0.005), which outperforms several existing deep learning methods. This study contributes to ongoing efforts to use deep learning models for automatically learning histologic patterns from whole slide images (WSIs) and predicting LUAD recurrence risk, thereby improving the accuracy and efficiency of treatment decision making.

RESUMEN

PURPOSE: Anasarca is a condition that results from organ dysfunctions, such as heart, kidney, or liver failure, characterized by the presence of edema throughout the body. The quantification of accumulated edema may have potential clinical benefits. This work focuses on accurately estimating the amount of edema non-invasively using abdominal CT scans, with minimal false positives. However, edema segmentation is challenging due to the complex appearance of edema and the lack of manually annotated volumes. METHODS: We propose a weakly supervised approach for edema segmentation using initial edema labels from the current state-of-the-art method for edema segmentation (Intensity Prior), along with labels of surrounding tissues as anatomical priors. A multi-class 3D nnU-Net was employed as the segmentation network, and training was performed using an iterative annotation workflow. RESULTS: We evaluated segmentation accuracy on a test set of 25 patients with edema. The average Dice Similarity Coefficient of the proposed method was similar to Intensity Prior (61.5% vs. 61.7%; p = 0.83 ). However, the proposed method reduced the average False Positive Rate significantly, from 1.8% to 1.1% ( p < 0.001 ). Edema volumes computed using automated segmentation had a strong correlation with manual annotation ( R 2 = 0.87 ). CONCLUSION: Weakly supervised learning using 3D multi-class labels and iterative annotation is an efficient way to perform high-quality edema segmentation with minimal false positives. Automated edema segmentation can produce edema volume estimates that are highly correlated with manual annotation. The proposed approach is promising for clinical applications to monitor anasarca using estimated edema volumes.

RESUMEN

Introduction: Pathologists rely on whole slide images (WSIs) to diagnose cancer by identifying tumor cells and subtypes. Deep learning models, particularly weakly supervised ones, classify WSIs using image tiles but may overlook false positives and negatives due to the heterogeneous nature of tumors. Both cancerous and healthy cells can proliferate in patterns that extend beyond individual tiles, leading to errors at the tile level that result in inaccurate tumor-level classifications. Methods: To address this limitation, we introduce NATMIL (Neighborhood Attention Transformer Multiple Instance Learning), which utilizes the Neighborhood Attention Transformer to incorporate contextual dependencies among WSI tiles. NATMIL enhances multiple instance learning by integrating a broader tissue context into the model. Our approach enhances the accuracy of tumor classification by considering the broader tissue context, thus reducing errors associated with isolated tile analysis. Results: We conducted a quantitative analysis to evaluate NATMIL's performance against other weakly supervised algorithms. When applied to subtyping non-small cell lung cancer (NSCLC) and lymph node (LN) tumors, NATMIL demonstrated superior accuracy. Specifically, NATMIL achieved accuracy values of 89.6% on the Camelyon dataset and 88.1% on the TCGA-LUSC dataset, outperforming existing methods. These results underscore NATMIL's potential as a robust tool for improving the precision of cancer diagnosis using WSIs. Discussion: Our findings demonstrate that NATMIL significantly improves tumor classification accuracy by reducing errors associated with isolated tile analysis. The integration of contextual dependencies enhances the precision of cancer diagnosis using WSIs, highlighting NATMILs´ potential as a robust tool in pathology.

RESUMEN

BACKGROUND: Recent advancements in anomaly detection have paved the way for novel radiological reading assistance tools that support the identification of findings, aimed at saving time. The clinical adoption of such applications requires a low rate of false positives while maintaining high sensitivity. PURPOSE: In light of recent interest and development in multi pathology identification, we present a novel method, based on a recent contrastive self-supervised approach, for multiple chest-related abnormality identification including low lung density area ("LLDA"), consolidation ("CONS"), nodules ("NOD") and interstitial pattern ("IP"). Our approach alerts radiologists about abnormal regions within a computed tomography (CT) scan by providing 3D localization. METHODS: We introduce a new method for the classification and localization of multiple chest pathologies in 3D Chest CT scans. Our goal is to distinguish four common chest-related abnormalities: "LLDA", "CONS", "NOD", "IP" and "NORMAL". This method is based on a 3D patch-based classifier with a Resnet backbone encoder pretrained leveraging recent contrastive self supervised approach and a fine-tuned classification head. We leverage the SimCLR contrastive framework for pretraining on an unannotated dataset of randomly selected patches and we then fine-tune it on a labeled dataset. During inference, this classifier generates probability maps for each abnormality across the CT volume, which are aggregated to produce a multi-label patient-level prediction. We compare different training strategies, including random initialization, ImageNet weight initialization, frozen SimCLR pretrained weights and fine-tuned SimCLR pretrained weights. Each training strategy is evaluated on a validation set for hyperparameter selection and tested on a test set. Additionally, we explore the fine-tuned SimCLR pretrained classifier for 3D pathology localization and conduct qualitative evaluation. RESULTS: Validated on 111 chest scans for hyperparameter selection and subsequently tested on 251 chest scans with multi-abnormalities, our method achieves an AUROC of 0.931 (95% confidence interval [CI]: [0.9034, 0.9557], p $ p$ -value < 0.001) and 0.963 (95% CI: [0.952, 0.976], p $ p$ -value < 0.001) in the multi-label and binary (i.e., normal versus abnormal) settings, respectively. Notably, our method surpasses the area under the receiver operating characteristic (AUROC) threshold of 0.9 for two abnormalities: IP (0.974) and LLDA (0.952), while achieving values of 0.853 and 0.791 for NOD and CONS, respectively. Furthermore, our results highlight the superiority of incorporating contrastive pretraining within the patch classifier, outperforming Imagenet pretraining weights and non-pretrained counterparts with uninitialized weights (F1 score = 0.943, 0.792, and 0.677 respectively). Qualitatively, the method achieved a satisfactory 88.8% completeness rate in localization and maintained an 88.3% accuracy rate against false positives. CONCLUSIONS: The proposed method integrates self-supervised learning algorithms for pretraining, utilizes a patch-based approach for 3D pathology localization and develops an aggregation method for multi-label prediction at patient-level. It shows promise in efficiently detecting and localizing multiple anomalies within a single scan.

RESUMEN

High performance of deep learning on medical image segmentation rely on large-scale pixel-level dense annotations, which poses a substantial burden on medical experts due to the laborious and time-consuming annotation process, particularly for 3D images. To reduce the labeling cost as well as maintain relatively satisfactory segmentation performance, weakly-supervised learning with sparse labels has attained increasing attentions. In this work, we present a scribble-based framework for medical image segmentation, called Dynamically Mixed Soft Pseudo-label Supervision (DMSPS). Concretely, we extend a backbone with an auxiliary decoder to form a dual-branch network to enhance the feature capture capability of the shared encoder. Considering that most pixels do not have labels and hard pseudo-labels tend to be over-confident to result in poor segmentation, we propose to use soft pseudo-labels generated by dynamically mixing the decoders' predictions as auxiliary supervision. To further enhance the model's performance, we adopt a two-stage approach where the sparse scribbles are expanded based on predictions with low uncertainties from the first-stage model, leading to more annotated pixels to train the second-stage model. Experiments on ACDC dataset for cardiac structure segmentation, WORD dataset for 3D abdominal organ segmentation and BraTS2020 dataset for 3D brain tumor segmentation showed that: (1) compared with the baseline, our method improved the average DSC from 50.46% to 89.51%, from 75.46% to 87.56% and from 52.61% to 76.53% on the three datasets, respectively; (2) DMSPS achieved better performance than five state-of-the-art scribble-supervised segmentation methods, and is generalizable to different segmentation backbones. The code is available online at: https://github.com/HiLab-git/DMSPS.

Asunto(s)

RESUMEN

Precise segmentation for skin cancer lesions at different stages is conducive to early detection and further treatment. Considering the huge cost of obtaining pixel-perfect annotations for this task, segmentation using less expensive image-level labels has become a research direction. Most image-level label weakly supervised segmentation uses class activation mapping (CAM) methods. A common consequence of this method is incomplete foreground segmentation, insufficient segmentation, or false negatives. At the same time, when performing weakly supervised segmentation of skin cancer lesions, ulcers, redness, and swelling may appear near the segmented areas of individual disease categories. This co-occurrence problem affects the model's accuracy in segmenting class-related tissue boundaries to a certain extent. The above two issues are determined by the loosely constrained nature of image-level labels that penalize the entire image space. Therefore, providing pixel-level constraints for weak supervision of image-level labels is the key to improving performance. To solve the above problems, this paper proposes a joint unsupervised constraint-assisted weakly supervised segmentation model (UCA-WSS). The weakly supervised part of the model adopts a dual-branch adversarial erasure mechanism to generate higher-quality CAM. The unsupervised part uses contrastive learning and clustering algorithms to generate foreground labels and fine boundary labels to assist segmentation and solve common co-occurrence problems in weakly supervised skin cancer lesion segmentation through unsupervised constraints. The model proposed in the article is evaluated comparatively with other related models on some public dermatology data sets. Experimental results show that our model performs better on the skin cancer segmentation task than other weakly supervised segmentation models, showing the potential of combining unsupervised constraint methods on weakly supervised segmentation.

Asunto(s)

RESUMEN

Despite that deep learning has achieved state-of-the-art performance for automatic medical image segmentation, it often requires a large amount of pixel-level manual annotations for training. Obtaining these high-quality annotations is time-consuming and requires specialized knowledge, which hinders the widespread application that relies on such annotations to train a model with good segmentation performance. Using scribble annotations can substantially reduce the annotation cost, but often leads to poor segmentation performance due to insufficient supervision. In this work, we propose a novel framework named as ScribSD+ that is based on multi-scale knowledge distillation and class-wise contrastive regularization for learning from scribble annotations. For a student network supervised by scribbles and the teacher based on Exponential Moving Average (EMA), we first introduce multi-scale prediction-level Knowledge Distillation (KD) that leverages soft predictions of the teacher network to supervise the student at multiple scales, and then propose class-wise contrastive regularization which encourages feature similarity within the same class and dissimilarity across different classes, thereby effectively improving the segmentation performance of the student network. Experimental results on the ACDC dataset for heart structure segmentation and a fetal MRI dataset for placenta and fetal brain segmentation demonstrate that our method significantly improves the student's performance and outperforms five state-of-the-art scribble-supervised learning methods. Consequently, the method has a potential for reducing the annotation cost in developing deep learning models for clinical diagnosis.

Asunto(s)

RESUMEN

Interstitial lung disease (ILD) is characterized by progressive pathological changes that require timely and accurate diagnosis. The early detection and progression assessment of ILD are important for effective management. This study introduces a novel quantitative evaluation method utilizing chest radiographs to analyze pixel-wise changes in ILD. Using a weakly supervised learning framework, the approach incorporates the contrastive unpaired translation model and a newly developed ILD extent scoring algorithm for more precise and objective quantification of disease changes than conventional visual assessments. The ILD extent score calculated through this method demonstrated a classification accuracy of 92.98% between ILD and normal classes. Additionally, using an ILD follow-up dataset for interval change analysis, this method assessed disease progression with an accuracy of 85.29%. These findings validate the reliability of the ILD extent score as a tool for ILD monitoring. The results of this study suggest that the proposed quantitative method may improve the monitoring and management of ILD.

RESUMEN

Purpose: The molecular classification of breast cancer is crucial for effective treatment. The emergence of digital pathology has ushered in a new era in which weakly supervised learning leveraging whole-slide images has gained prominence in developing deep learning models because this approach alleviates the need for extensive manual annotation. Weakly supervised learning was employed to classify the molecular subtypes of breast cancer. Methods: Our approach capitalizes on two whole-slide image datasets: one consisting of breast cancer cases from the Korea University Guro Hospital (KG) and the other originating from The Cancer Genomic Atlas dataset (TCGA). Furthermore, we visualized the inferred results using an attention-based heat map and reviewed the histomorphological features of the most attentive patches. Results: The KG+TCGA-trained model achieved an area under the receiver operating characteristics value of 0.749. An inherent challenge lies in the imbalance among subtypes. Additionally, discrepancies between the two datasets resulted in different molecular subtype proportions. To mitigate this imbalance, we merged the two datasets, and the resulting model exhibited improved performance. The attentive patches correlated well with widely recognized histomorphologic features. The triple-negative subtype has a high incidence of high-grade nuclei, tumor necrosis, and intratumoral tumor-infiltrating lymphocytes. The luminal A subtype showed a high incidence of collagen fibers. Conclusions: The artificial intelligence (AI) model based on weakly supervised learning showed promising performance. A review of the most attentive patches provided insights into the predictions of the AI model. AI models can become invaluable screening tools that reduce costs and workloads in practice.

RESUMEN

The annotation of magnetic resonance imaging (MRI) images plays an important role in deep learning-based MRI segmentation tasks. Semi-automatic annotation algorithms are helpful for improving the efficiency and reducing the difficulty of MRI image annotation. However, the existing semi-automatic annotation algorithms based on deep learning have poor pre-annotation performance in the case of insufficient segmentation labels. In this paper, we propose a semi-automatic MRI annotation algorithm based on semi-weakly supervised learning. In order to achieve a better pre-annotation performance in the case of insufficient segmentation labels, semi-supervised and weakly supervised learning were introduced, and a semi-weakly supervised learning segmentation algorithm based on sparse labels was proposed. In addition, in order to improve the contribution rate of a single segmentation label to the performance of the pre-annotation model, an iterative annotation strategy based on active learning was designed. The experimental results on public MRI datasets show that the proposed algorithm achieved an equivalent pre-annotation performance when the number of segmentation labels was much less than that of the fully supervised learning algorithm, which proves the effectiveness of the proposed algorithm.

RESUMEN

Instrument pose estimation is a key demand in computer-aided surgery, and its main challenges lie in two aspects: Firstly, the difficulty of obtaining stable corresponding image feature points due to the instruments' high refraction and complicated background, and secondly, the lack of labeled pose data. This study aims to tackle the pose estimation problem of surgical instruments in the current endoscope system using a single endoscopic image. More specifically, a weakly supervised method based on the instrument's image segmentation contour is proposed, with the effective assistance of synthesized endoscopic images. Our method consists of the following three modules: a segmentation module to automatically detect the instrument in the input image, followed by a point inference module to predict the image locations of the implicit feature points of the instrument, and a point back-propagatable Perspective-n-Point module to estimate the pose from the tentative 2D-3D corresponding points. To alleviate the over-reliance on point correspondence accuracy, the local errors of feature point matching and the global inconsistency of the corresponding contours are simultaneously minimized. Our proposed method is validated with both real and synthetic images in comparison with the current state-of-the-art methods.

RESUMEN

Colorectal cancer (CRC) is the third most common cancer in the United States. Tumor Budding (TB) detection and quantification are crucial yet labor-intensive steps in determining the CRC stage through the analysis of histopathology images. To help with this process, we adapt the Segment Anything Model (SAM) on the CRC histopathology images to segment TBs using SAM-Adapter. In this approach, we automatically take task-specific prompts from CRC images and train the SAM model in a parameter-efficient way. We compare the predictions of our model with the predictions from a trained-from-scratch model using the annotations from a pathologist. As a result, our model achieves an intersection over union (IoU) of 0.65 and an instance-level Dice score of 0.75, which are promising in matching the pathologist's TB annotation. We believe our study offers a novel solution to identify TBs on H&E-stained histopathology images. Our study also demonstrates the value of adapting the foundation model for pathology image segmentation tasks.

RESUMEN

BACKGROUND AND OBJECTIVE: Current automatic electrocardiogram (ECG) diagnostic systems could provide classification outcomes but often lack explanations for these results. This limitation hampers their application in clinical diagnoses. Previous supervised learning could not highlight abnormal segmentation output accurately enough for clinical application without manual labeling of large ECG datasets. METHOD: In this study, we present a multi-instance learning framework called MA-MIL, which has designed a multi-layer and multi-instance structure that is aggregated step by step at different scales. We evaluated our method using the public MIT-BIH dataset and our private dataset. RESULTS: The results show that our model performed well in both ECG classification output and heartbeat level, sub-heartbeat level abnormal segment detection, with accuracy and F1 scores of 0.987 and 0.986 for ECG classification and 0.968 and 0.949 for heartbeat level abnormal detection, respectively. Compared to visualization methods, the IoU values of MA-MIL improved by at least 17 % and at most 31 % across all categories. CONCLUSIONS: MA-MIL could accurately locate the abnormal ECG segment, offering more trustworthy results for clinical application.

Asunto(s)

RESUMEN

Deep learning plays a significant role in the detection of pulmonary nodules in low-dose computed tomography (LDCT) scans, contributing to the diagnosis and treatment of lung cancer. Nevertheless, its effectiveness often relies on the availability of extensive, meticulously annotated dataset. In this paper, we explore the utilization of an incompletely annotated dataset for pulmonary nodules detection and introduce the FULFIL (Forecasting Uncompleted Labels For Inexpensive Lung nodule detection) algorithm as an innovative approach. By instructing annotators to label only the nodules they are most confident about, without requiring complete coverage, we can substantially reduce annotation costs. Nevertheless, this approach results in an incompletely annotated dataset, which presents challenges when training deep learning models. Within the FULFIL algorithm, we employ Graph Convolution Network (GCN) to discover the relationships between annotated and unannotated nodules for self-adaptively completing the annotation. Meanwhile, a teacher-student framework is employed for self-adaptive learning using the completed annotation dataset. Furthermore, we have designed a Dual-Views loss to leverage different data perspectives, aiding the model in acquiring robust features and enhancing generalization. We carried out experiments using the LUng Nodule Analysis (LUNA) dataset, achieving a sensitivity of 0.574 at a False positives per scan (FPs/scan) of 0.125 with only 10% instance-level annotations for nodules. This performance outperformed comparative methods by 7.00%. Experimental comparisons were conducted to evaluate the performance of our model and human experts on test dataset. The results demonstrate that our model can achieve a comparable level of performance to that of human experts. The comprehensive experimental results demonstrate that FULFIL can effectively leverage an incomplete pulmonary nodule dataset to develop a robust deep learning model, making it a promising tool for assisting in lung nodule detection.

Asunto(s)

RESUMEN

Weakly supervised temporal action localization aims to locate the temporal boundaries of action instances in untrimmed videos using video-level labels and assign them the corresponding action category. Generally, it is solved by a pipeline called "localization-by-classification", which finds the action instances by classifying video snippets. However, since this approach optimizes the video-level classification objective, the generated activation sequences often suffer interference from class-related scenes, resulting in a large number of false positives in the prediction results. Many existing works treat background as an independent category, forcing models to learn to distinguish background snippets. However, under weakly supervised conditions, the background information is fuzzy and uncertain, making this method extremely difficult. To alleviate the impact of false positives, we propose a new actionness-guided false positive suppression framework. Our method seeks to suppress false positive backgrounds without introducing the background category. Firstly, we propose a self-training actionness branch to learn class-agnostic actionness, which can minimize the interference of class-related scene information by ignoring the video labels. Secondly, we propose a false positive suppression module to mine false positive snippets and suppress them. Finally, we introduce the foreground enhancement module, which guides the model to learn the foreground with the help of the attention mechanism as well as class-agnostic actionness. We conduct extensive experiments on three benchmarks (THUMOS14, ActivityNet1.2, and ActivityNet1.3). The results demonstrate the effectiveness of our method in suppressing false positives and it achieves the state-of-the-art performance. Code: https://github.com/lizhilin-ustc/AFPS.

Asunto(s)

RESUMEN

Automatic crack segmentation plays an essential role in maintaining the structural health of buildings and infrastructure. Despite the success in fully supervised crack segmentation, the costly pixel-level annotation restricts its application, leading to increased exploration in weakly supervised crack segmentation (WSCS). However, WSCS methods inevitably bring in noisy pseudo-labels, which results in large fluctuations. To address this problem, we propose a novel confidence-aware co-training (CAC) framework for WSCS. This framework aims to iteratively refine pseudo-labels, facilitating the learning of a more robust segmentation model. Specifically, a co-training mechanism is designed and constructs two collaborative networks to learn uncertain crack pixels, from easy to hard. Moreover, the dynamic division strategy is designed to divide the pseudo-labels based on the crack confidence score. Among them, the high-confidence pseudo-labels are utilized to optimize the initialization parameters for the collaborative network, while low-confidence pseudo-labels enrich the diversity of crack samples. Extensive experiments conducted on the Crack500, DeepCrack, and CFD datasets demonstrate that the proposed CAC significantly outperforms other WSCS methods.

RESUMEN

Peripancreatic vessel segmentation and anatomical labeling are pivotal aspects in aiding surgical planning and prognosis for patients with pancreatic tumors. Nevertheless, prevailing techniques often fall short in achieving satisfactory segmentation performance for the peripancreatic vein (PPV), leading to predictions characterized by poor integrity and connectivity. Besides, unsupervised labeling algorithms usually cannot deal with complex anatomical variation while fully supervised methods require a large number of voxel-wise annotations for training, which is very labor-intensive and time-consuming. To address these two problems, we propose an Automated Peripancreatic vEssel Segmentation and lAbeling (APESA) framework, to not only highly improve the segmentation performance for PPV, but also efficiently identify the peripancreatic artery (PPA) branches. There are two core modules in our proposed APESA framework: iterative trunk growth module (ITGM) for vein segmentation and weakly supervised labeling mechanism (WSLM) for artery labeling. The ITGM is composed of a series of iterative submodules, each of which chooses the largest connected component of the previous PPV segmentation as the trunk of a tree structure, seeks for the potential missing branches around the trunk by our designed branch proposal network, and facilitates trunk growth under the connectivity constraint. The WSLM incorporates the rule-based pseudo label generation with less expert participation, an anatomical labeling network to learn the branch distribution voxel by voxel, and adaptive radius-based postprocessing to refine the branch structures of the labeling predictions. Our achieved Dice of 94.01% for PPV segmentation on our collected dataset represents an approximately 10% accuracy improvement compared to state-of-the-art methods. Additionally, we attained a Dice of 97.01% for PPA segmentation and competitive labeling performance for PPA labeling compared to prior works. Our source codes will be publicly available at https://github.com/ZouLiwen-1999/APESA.

Asunto(s)

RESUMEN

Fringe projection profilometry (FPP) is widely used for high-accuracy 3D imaging. However, employing multiple sets of fringe patterns ensures 3D reconstruction accuracy while inevitably constraining the measurement speed. Conventional dual-frequency FPP reduces the number of fringe patterns for one reconstruction to six or fewer, but the highest period-number of fringe patterns generally is limited because of phase errors. Deep learning makes depth estimation from fringe images possible. Inspired by unsupervised monocular depth estimation, this paper proposes a novel, weakly supervised method of depth estimation for single-camera FPP. The trained network can estimate the depth from three frames of 64-period fringe images. The proposed method is more efficient in terms of fringe pattern efficiency by at least 50% compared to conventional FPP. The experimental results show that the method achieves competitive accuracy compared to the supervised method and is significantly superior to the conventional dual-frequency methods.

RESUMEN

The value of coarsely labeled datasets in learning transferable representations for medical images is investigated in this work. Compared to fine labels which require meticulous effort to annotate, coarse labels can be acquired at a significantly lower cost and can provide useful training signals for data-hungry deep neural networks. We consider coarse labels in the form of binary labels differentiating a normal (healthy) image from an abnormal (diseased) image and propose CAMContrast, a two-stage representation learning framework for medical images. Using class activation maps, CAMContrast makes use of the binary labels to generate heatmaps as positive views for contrastive representation learning. Specifically, the learning objective is optimized to maximize the agreement within fixed crops of image-heatmap pair to learn fine-grained representations that are generalizable to different downstream tasks. We empirically validate the transfer learning performance of CAMContrast on several public datasets, covering classification and segmentation tasks on fundus photographs and chest X-ray images. The experimental results showed that our method outperforms other self-supervised and supervised pretrain methods in terms of data efficiency and downstream performance.

Asunto(s)

RESUMEN

BACKGROUND: Automated estimation of Pulmonary function test (PFT) results from Computed Tomography (CT) could advance the use of CT in screening, diagnosis, and staging of restrictive pulmonary diseases. Estimating lung function per lobe, which cannot be done with PFTs, would be helpful for risk assessment for pulmonary resection surgery and bronchoscopic lung volume reduction. PURPOSE: To automatically estimate PFT results from CT and furthermore disentangle the individual contribution of pulmonary lobes to a patient's lung function. METHODS: We propose I3Dr, a deep learning architecture for estimating global measures from an image that can also estimate the contributions of individual parts of the image to this global measure. We apply it to estimate the separate contributions of each pulmonary lobe to a patient's total lung function from CT, while requiring only CT scans and patient level lung function measurements for training. I3Dr consists of a lobe-level and a patient-level model. The lobe-level model extracts all anatomical pulmonary lobes from a CT scan and processes them in parallel to produce lobe level lung function estimates that sum up to a patient level estimate. The patient-level model directly estimates patient level lung function from a CT scan and is used to re-scale the output of the lobe-level model to increase performance. After demonstrating the viability of the proposed approach, the I3Dr model is trained and evaluated for PFT result estimation using a large data set of 8 433 CT volumes for training, 1 775 CT volumes for validation, and 1 873 CT volumes for testing. RESULTS: First, we demonstrate the viability of our approach by showing that a model trained with a collection of digit images to estimate their sum implicitly learns to assign correct values to individual digits. Next, we show that our models can estimate lobe-level quantities, such as COVID-19 severity scores, pulmonary volume (PV), and functional pulmonary volume (FPV) from CT while only provided with patient-level quantities during training. Lastly, we train and evaluate models for producing spirometry and diffusion capacity of carbon mono-oxide (DLCO) estimates at the patient and lobe level. For producing Forced Expiratory Volume in one second (FEV1), Forced Vital Capacity (FVC), and DLCO estimates, I3Dr obtains mean absolute errors (MAE) of 0.377 L, 0.297 L, and 2.800 mL/min/mm Hg respectively. We release the resulting algorithms for lung function estimation to the research community at https://grand-challenge.org/algorithms/lobe-wise-lung-function-estimation/ CONCLUSIONS: I3Dr can estimate global measures from an image, as well as the contributions of individual parts of the image to this global measure. It offers a promising approach for estimating PFT results from CT scans and disentangling the individual contribution of pulmonary lobes to a patient's lung function. The findings presented in this work may advance the use of CT in screening, diagnosis, and staging of restrictive pulmonary diseases as well as in risk assessment for pulmonary resection surgery and bronchoscopic lung volume reduction.

Asunto(s)